System and method for detection of gas in the atmosphere from a multispectral image

Abstract

A system and method for detection of gas in the atmosphere front a multispectral image including a plurality of pixels, each pixel having a spectral signature including a set of intensities of electromagnetic (EM) energy reflected at various bands, the method including: clustering the plurality of pixels based on channels of the spectral signatures that are indicative of presence of the gas, to produce a first set of clusters; clustering the plurality of pixels based on channels of the spectral signatures that are non-indicative of presence of the gas, to produce a second set of clusters; matching clusters from the first set of clusters and the second set of clusters; and labeling clusters that are present in the first set of clusters and not present in the second set of clusters as suspected as including the gas.

The invention claimed is: 
     
       1. A method for detection of gas in the atmosphere from a multispectral image comprising a plurality of pixels, each pixel having a spectral signature comprising a set of intensities of electromagnetic (EM) energy reflected at various bands, the method comprising:
 clustering the plurality of pixels based on channels of the spectral signatures that are indicative of presence of the gas, to produce a first set of clusters; 
 clustering the plurality of pixels based on channels of the spectral signatures that are non-indicative of presence of the gas, to produce a second set of clusters; 
 matching clusters from the first set of clusters and the second set of clusters; 
 for each pixel in the multispectral image, performing a mathematical operation between at least one indicative channel and at least one non-indicative channel, to generate a derived image; 
 clustering the plurality of pixels in the derived image, to produce a third set of clusters; 
 calculating a unified derived value for each cluster in the third set of clusters; 
 unifying clusters that are present in the first set of clusters and not present in the second set of clusters with clusters in the third set of clusters with the unified derived value above the threshold; and 
 labeling the unified clusters as suspected as including the gas. 
 
     
     
       2. The method of  claim 1 , further comprising:
 for each of the labeled clusters: calculating a difference between the non-indicative channels of the labeled cluster with corresponding channels in each of the clusters in the second set of clusters; and 
 concluding that a background material in the labeled cluster is similar to a background material in the cluster in the second set of clusters if the difference is below a threshold, and that the background material is different otherwise. 
 
     
     
       3. The method of  claim 1 , wherein matching clusters from the first set of clusters and the second set of clusters is performed based on geometrical or spatial parameters of the first set and the second set of clusters. 
     
     
       4. The method of  claim 1 , comprising:
 merging the labeled clusters based on Euclidian distance and trend. 
 
     
     
       5. The method of  claim 1 , comprising:
 generating a convex hull around the labeled clusters, and labeling the area inside the convex hull as suspected as including the gas. 
 
     
     
       6. The method of  claim 1 , further comprising:
 calculating a unified spectral signature for each cluster in the first set of clusters; 
 for each pair of clusters in the first set of clusters: 
 dividing each value in a unified spectral signature of a first cluster by a corresponding value of in a unified spectral signature of a second cluster, to obtain a set of ratios for the two clusters; 
 determining that a cluster of the pair of clusters is suspected as including gas based on the ratios; and 
 unifying the labeled clusters with the clusters that are suspected as including gas based on the ratios. 
 
     
     
       7. The method of  claim 6 , wherein unifying comprises performing a logical operation between the labeled clusters and the clusters that are suspected as including gas based on the ratios. 
     
     
       8. The method of  claim 6 , wherein determining that a cluster of the pair of clusters is suspected as including gas based on the ratios comprises:
 determining that a cluster of the pair of clusters is suspected as including gas if an average of the ratios in the indicative channels minus an average of the ratios in the nonindicative channels is larger than a standard deviation of the ratios in the nonindicative channels multiplied by a factor equal or larger than 1. 
 
     
     
       9. The method of  claim 1 , wherein unifying comprises performing a logical operation between the labeled clusters and the clusters with the unified derived value above the threshold. 
     
     
       10. The method of  claim 1 , further comprising:
 presenting the multispectral image with a visual marking of the clusters that are suspected as including the gas. 
 
     
     
       11. A method for detection of gas in the atmosphere from a multispectral image comprising a plurality of pixels, each having a spectral signature comprising a vector of intensities of electromagnetic (EM) energy reflected at various wavelength bands, the method comprising:
 clustering the plurality of pixels based on channels of the spectral signatures that are known to be indicative of presence of the gas, to produce a set of clusters; 
 calculating a unified spectral signature for each cluster in the set of clusters; 
 for each pair of clusters in the set of clusters:
 dividing each value in a first unified spectral signature by a corresponding value of in a first unified spectral signature, to obtain a set of ratios; and 
 determining that a cluster of the pair of clusters is suspected as including gas if an average of the ratios in the indicative channels minus an average of the ratios in the nonindicative channels is larger than a standard deviation of the ratios in the nonindicative channels multiplied by a factor equal or larger than 1. 
 
 
     
     
       12. The method of  claim 11 , comprising:
 merging the clusters that are determined as including gas based on Euclidian distance and trend. 
 
     
     
       13. The method of  claim 11 , comprising:
 generating a convex hull around the clusters that are determined as including gas, and labeling the area inside the convex hull as suspected as including the gas. 
 
     
     
       14. A system for detection of gas in the atmosphere from a multispectral image comprising a plurality of pixels, each pixel having a spectral signature comprising a set of intensities of electromagnetic (EM) energy reflected at various bands, the system comprising:
 a memory; 
 a processor configured to: 
 cluster the plurality of pixels based on channels of the spectral signatures that are indicative of presence of the gas, to produce a first set of clusters; 
 cluster the plurality of pixels based on channels of the spectral signatures that are non-indicative of presence of the gas, to produce a second set of clusters; 
 match clusters from the first set of clusters and the second set of clusters; 
 label clusters that are present in the first set of clusters and not present in the second set of clusters as suspected as including the gas, 
 calculate a unified spectral signature for each cluster in the first set of clusters; 
 for each pair of clusters in the first set of clusters:
 divide each value in a unified spectral signature of a first cluster by a corresponding value of in a unified spectral signature of a second cluster, to obtain a set of ratios for the two clusters; 
 determine that a cluster of the pair of clusters is suspected as including gas based on the ratios; and 
 
 unify the labeled clusters with the clusters that are suspected as including gas based on the ratios. 
 
     
     
       15. The system of  claim 14 , wherein the processor is further configured to:
 for each of the unified clusters: calculate a difference between the non-indicative channels of the cluster with corresponding channels in each of the clusters in the second set of clusters; and 
 conclude that a background material in the unified cluster is similar to a background material in the cluster in the second set of clusters if the difference is below a threshold, and that the background material is different otherwise. 
 
     
     
       16. The system of  claim 14 , wherein the processor is configured to match clusters from the first set of clusters and the second set of clusters based on geometrical or spatial parameters of the first set and the second set of clusters. 
     
     
       17. The system of  claim 14 , wherein the processor is further configured to:
 merge the labeled unified clusters based on Euclidian distance and trend; 
 generate a convex hull around the merged clusters; and 
 label the area inside the convex hull as suspected as including the gas. 
 
     
     
       18. The system of  claim 14 , wherein the processor is configured to unify the labeled clusters with the clusters that are suspected as including gas by performing a logical operation between the labeled clusters and the clusters that are suspected as including gas based on the ratios. 
     
     
       19. The system of  claim 18 , wherein the processor is configured to determine that a cluster of the pair of clusters is suspected as including gas based on the ratios by:
 determining that a cluster of the pair of clusters is suspected as including gas if an average of the ratios in the indicative channels minus an average of the ratios in the nonindicative channels is larger than a standard deviation of the ratios in the nonindicative channels multiplied by a factor equal or larger than 1. 
 
     
     
       20. The system of  claim 14 , further comprising:
 for each pixel in the multispectral image, performing a mathematical operation between at least one indicative channel and at least one non-indicative channel, to generate a derived image; 
 clustering the plurality of pixels in the derived image, to produce a third set of clusters; 
 calculating a unified derived value for each cluster in the third set of clusters; 
 labeling clusters in the third set of clusters with a unified derived value above a threshold as suspected as including the gas; and 
 unifying the unified clusters with the clusters with the unified derived value above the threshold.